Infants' ability to track temporarily occluded objects that moved on circular trajectories was investigated in 20 infants using a longitudinal design. They were first seen at 6 months and then every 2nd month until the end of their 1st year. Infants were presented with occlusion events covering 20% of the target's trajectory (effective occlusion interval ranged from 500–4,000 msec). Gaze was measured using an ASL 504 infrared eye-tracking system. Results effectively demonstrate that infants from 6 months of age can represent the spatiotemporal dynamics of occluded objects. Infants at all ages tested were able to predict, under certain conditions, when and where the object would reappear after occlusion. They moved gaze accurately to the position where the object was going to reappear and scaled their timing to the current occlusion duration. The average rate of predictive gaze crossings increased with occlusion duration. These results are discussed as a 2-factor process. Successful predictions are dependent on strong representations, themselves dependent on the richness of information available during encoding and graded representations.